Abstract: This research was aimed to find out: (1) which one gave better mathematics learning achievement, learning model TS-TS with Realistic Mathematics Education (RME) approach, TPS with RME approach or conventional learning; (2) which one had better mathematics learning achievement, students having high, medium or low mathematics learning activity; (3) in each studentsâ levels of mathematics learning activity which one gave better mathematics learning achievement, learning model TS-TS with RME approach, TPS with RME approach or conventional learning; (4) in each learning model, the levels of mathematics learning activity which one had better mathematics learning achievement, students having high, medium or low mathematics learning activity. The research was a quasi experimental research with 3x3 factorial design. The population of research was all students of Junior High School (SMP) in Karanganyar. The samples were chosen by using stratified cluster random sampling. The samples were 288 students. Pre-requisite tests were used Lilliefors method for normality test and Bartlett method for homogeneity test. After examining the data, it showed that the data had same variance and they were in normal distribution. Prior knowledge data was examined by using one-way analysis of variance with unbalanced cells. It showed that three populations had balance in prior knowledge. Meanwhile, the technique of analyzing the data was two-ways analysis of variance with unbalanced cells. The conclusions of this research showed as follows. (1) Learning model TS-TS with RME approach provided better mathematics learning achievement than TPS with RME approach and conventional learning, learning model TPS with RME provided better mathematics learning achievement than conventional learning. (2) The students having high mathematics learning activity had better mathematics learning achievement than the students having medium and low mathematics learning activity. The students having medium mathematics learning activity had better mathematics learning achievement than the students having low mathematics learning activity. (3) In each category of mathematics learning activity, learning model TS-TS with RME approach was better than learning model TPS with RME approach and conventional learning, learning model TPS with RME approach was better than conventional learning. (4) In each learning model, learning achievement of students with high mathematics learning activity was better than learning achievement of students with medium and low mathematics learning activity, learning achievement of students with medium mathematics learning activity was better than the learning achievement of students with low mathematics learning activity.Keywords: TS-TS with RME approach, TPS with RME approach, mathematics learning activity, mathematics learning achievement.